Assessing ozone formation impact through SHAP interaction redistribution analysis: A novel framework for evaluating VOC photochemical loss and source interactions

Abstract

Despite extensive research on tropospheric ozone (O₃) formation, quantifying the impact of photochemical processes on source contributions remains challenging. This study developed a SHAP interaction redistribution analysis (SIRA) framework that systematically analyzes and reallocates machine learning interpretation results to quantify photochemical impacts on O₃ formation and source contributions. Based on improved photochemical age parameterization method, the mean VOC photochemical loss concentration was 7.44 ± 10.36 ppb, with alkenes, aromatics, and OVOCs showing the highest ratios of photochemical loss to observed concentrations (1.35, 0.90, and 0.66, respectively). SHAP analysis revealed meteorological factors and NO as primary contributors to O₃ formation (50.82 % and 12.24 %, respectively). The importance ratio of TVOCs to NO_x was 1.5 during the study period and increased to 2.4 during O₃ pollution episodes, indicating VOC-limited O₃ formation with enhanced VOC sensitivity during pollution periods. The SIRA framework identified secondary generation (48.38 %), biogenic emissions (13.87 %), and industrial emissions (9.33 %) as major O₃ contributors under photochemical loss conditions. After redistributing secondary generation through source interaction analysis, biogenic emissions and biomass burning showed the highest increases in SHAP importance (41.18 % with mean |SHAP values| = 1.87 and 40.27 % with mean |SHAP values| = 0.98, respectively), identifying them as the primary precursor sources for secondary generation. Among anthropogenic sources, vehicle emissions became the dominant contributor with a 17.0 % increase in contribution. This framework offers an innovative approach for quantifying the photochemical impacts on source contributions, enabling more targeted and effective strategies for O₃ pollution control.